Recombinant production of the antibody fragment D1.3 scFv with different Bacillus strains

Background Different strains of the genus Bacillus are versatile candidates for the industrial production and secretion of heterologous proteins. They can be cultivated quite easily, show high growth rates and are usually non-pathogenic and free of endo- and exotoxins. They have the ability to secrete proteins with high efficiency into the growth medium, which allows cost-effective downstream purification processing. Some of the most interesting and challenging heterologous proteins are recombinant antibodies and antibody fragments. They are important and suitable tools in medical research for analytics, diagnostics and therapy. The smallest conventional antibody fragment with high-affinity binding to an antigen is the single-chain fragment variable (scFv). Here, different strains of the genus Bacillus were investigated using diverse cultivation systems for their suitability to produce and secret a recombinant scFv. Results Extracellular production of lysozyme-specific scFv D1.3 was realized by constructing a plasmid with a xylose-inducible promoter optimized for Bacillus megaterium and the D1.3scFv gene fused to the coding sequence of the LipA signal peptide from B. megaterium. Functional scFv was successfully secreted with B. megaterium MS941, Bacillus licheniformis MW3 and the three Bacillus subtilis strains 168, DB431 and WB800N differing in the number of produced proteases. Starting with shake flasks (150 mL), the bioprocess was scaled down to microtiter plates (1250 µL) as well as scaled up to laboratory-scale bioreactors (2 L). The highest extracellular concentration of D1.3 scFv (130 mg L−1) and highest space–time-yield (8 mg L−1 h−1) were accomplished with B. subtilis WB800N, a strain deficient in eight proteases. These results were reproduced by the production and secretion of a recombinant penicillin G acylase (Pac). Conclusions The genus Bacillus provides high potential microbial host systems for the secretion of challenging heterologous proteins like antibody fragments and large proteins at high titers. In this study, the highest extracellular concentration and space–time-yield of a recombinant antibody fragment for a Gram-positive bacterium so far was achieved. The successful interspecies use of the here-designed plasmid originally optimized for B. megaterium was demonstrated by two examples, an antibody fragment and a penicillin G acylase in up to five different Bacillus strains. Electronic supplementary material The online version of this article (doi:10.1186/s12934-017-0625-9) contains supplementary material, which is available to authorized users.

natural products, Bacillus strains are great candidates for industrial production and secretion of heterologous proteins due to several advantages. In comparison to eukaryotic systems, their cultivation is simple and their high growth rates lead to short cultivation times. Most Bacillus strains are non-pathogenic and free of exo-and endotoxins. Species like B. subtilis and B. licheniformis even have the generally regarded as safe (GRAS) status [10,11]. In addition, these species have the ability to secrete proteins directly into the extracellular medium, resulting in cost-effective downstream purification processing. In contrast, Gram-negative bacteria, like the best-analyzed representative Escherichia coli, accumulate proteins intracellularly and tend to build mostly cell-toxic and insoluble protein accumulations (inclusion bodies). These result in cost-intensive downstream processing due to incorrect protein folding and inefficient disulfide bridge formation [3,12].
The production of proteins of eukaryotic origin in prokaryotic organisms provides one of the biggest challenges of biotechnology. Eukaryotic proteins heterologously produced and secreted with Bacillus are e.g. human interferon alpha (15 mg L −1 , B. subtilis, [13]), growth hormone (200 mg L −1 , B. subtilis, [14]) or epidermal growth factor (240 mg L −1 , Bacillus brevis, [15]). Levels of these heterologous proteins are quite low in comparison to homologous proteins and those of prokaryotic (Gramnegative or -positive) origin, e.g. secreted α-amylase with up to 3 g L −1 [16]. One major limiting factor is the degradation of the secreted protein by membrane-bound, cell wall-associated or secreted proteases [17]. Proteases constitute the cellular quality control which degrade proteins folded too slowly or incorrectly [18]. Many heterologous proteins are frequently found to be inefficiently and slowly folded compared to the homologous counterparts since they are not coevolved with the host protein production machinery. This might even lead to the exposure of protease-recognition sequences even in correctly folded conformation [16,19,20]. To overcome degradation of secreted heterologous proteins, many protease-deficient Bacillus mutants have been successfully constructed and used resulting in more effective extracellular protein production [21].
Some of the most interesting and challenging proteins heterologously produced in prokaryotes are recombinant antibodies and antibody fragments which are important and suitable tools in research and medicine. Their specific antigen binding is used in analytics, in proteome research, in diagnostics of pathogens and toxins and in therapy of inflammatory and tumor diseases [22][23][24]. A whole immunoglobulin G molecule is a hetero-tetramer with two heavy and two light chains connected by disulfide bridges and intramolecular disulfide bridges for stabilization [25]. Since their production requires a complicated folding apparatus and an oxidizing environment for disulfide bridge formation, many microbial host systems fail to produce significant amounts of the molecules [26]. In addition, bacterial hosts usually do not accomplish the correct glycosylation of the produced antibodies. However, smaller and simpler antibody fragments with full antigen binding capacity have been developed for research purposes, where biological activity is more important than structural authenticity and correct glycosylation pattern. The smallest conventional antibody fragment with high-affinity binding to an antigen is the single-chain fragment variable (scFv) [27], a heterodimer comprising the antibody heavyand light-chain variable domains connected by a peptide linker and with natural disulfide bonds within the chains to stabilize the molecule [28].
Here, the recombinant production and secretion of the lysozyme-specific antibody fragment D1.3 scFv in B. megaterium, B. licheniformis and three B. subtilis strains differing in their protease equipment are presented to demonstrate the interspecies use of the B. megaterium plasmid system. For this, the gene encoding His-tagged D1.3 scFv genetically fused to a suitable secretion signal was expressed under the control of a xylose-inducible promoter optimized for recombinant protein production in B. megaterium. Starting with uncontrolled shake flask (150 mL culture volume) cultivations, the bioprocess was scaled down to microtiter plates (1250 µL culture volume), but also scaled up to controlled laboratory-scale stirred tank bioreactors (2 L culture volume). Growth, secretion behavior, product titer, product kinetics and space-time-yield (STY) were compared between the different strains and cultivation systems. Finally, these results were reproduced for the production and the secretion of a recombinant penicillin G acylase (Pac) with a three times higher molecular weight compared to the D1.3 scFv. Pac (EC 3.5.1.11) catalyzes the cleavage of penicillin G and cephalosporin G to intermediates important for the production of semi-synthetic β-lactam antibiotics [29]. It is autocatalytically processed from a single cytoplasmic polypeptide precursor carrying a signal peptide and a spacer peptide and finally correctly folded outside the cell before released to the growth medium [30,31]. Therefore, Pac is a challenging candidate for recombinant secretion carried out here with three different Bacillus strains using the same plasmid background as for the recombinant D1.3 scFv.

Plasmid for the recombinant production and secretion of D1.3 scFv
The xylose-inducible promoter P xylA was originally derived from the B. megaterium strain DSM319. There it controls the xylABT operon encoding proteins of xylose uptake and metabolism via a divergently encoded repressor protein, termed XylR. P xylA was systematically optimized for high yield protein production in B. megaterium previously [32]. For the production, secretion and purification of D1.3 scFv in B. megaterium, the corresponding synthetic gene was fused to the coding sequences for the signal peptide of the lipase A of B. megaterium and for a his 6 -tag under control of native P xylA before [33]. The employed vector backbone was carrying the oriU, repU and tetL genes from the B. cereus plasmid pBC16 [34] for replication and selection in Bacillus. To now enhance the recombinant production and secretion of D1.3 scFv, the D1.3scFv gene fused to the lipase A signal peptide coding sequence was cloned into a pBC16-derived plasmid carrying the P xylA , optimized for recombinant protein production in B. megaterium (Fig. 1). Furthermore, this new plasmid pRBBm117 is minimized by around 700 bp to reduce metabolic burden for the production strain [35].
After transformation of B. megaterium MS941, a mutant of DSM319 lacking the main extracellular protease NprM and thus around 98% of extracellular protease activity [36], the recombinant secretion of D1.3 scFv was analyzed in 150 mL shake flask cultivation in minimal medium ( Table 1). The oxygen input was realized via orbital shaking movements. The volumetric mass transfer coefficients for oxygen (k L a) of 30-50 h −1 were assumed [37,38]. Growth passed the typical phases of a batch process (Fig. 2). The addition of 5 g L −1 xylose for the induction of recombinant gene expression 2-3 h after inoculation led to temporarily decreased growth. The specific growth rate µ [h −1 ] is the proportionality factor between time-dependent change of cell concentration dX/dt [g L −1 h −1 ] and current cell concentration X [g L −1 ] (dX/dt = µ · X). It decreased from 0.45 to 0.15 h −1 . Growth became accelerated again (0.33 h −1 ) until the initial 5 g L −1 fructose were completely consumed after 16-18 h. Stationary phase was attained with about 2 g L −1 of cell dry weight. The secretion of active, correctly folded D1.3 scFv already started during the exponential growth phase while maximal amounts of 8 mg L −1 were detected 4 h after entering the stationary phase and stayed constant until the end of the cultivation. The production was assumed mixed-growthassociated: the growth-associated biomass-related production yield coefficient α ≡ Y P/X = 0.13 mg scFv g −1 cells and the non-growth-associated specific production rate β = 0.41 mg scFv g −1 cells h −1 were calculated (Additional file 1: Table S1). During stationary phase, the carbon source fructose was already depleted and the formerly secreted acetate was consumed. The concentration of the inducer xylose stayed constant at a level of 5 g L −1 until fructose was exhausted and was only slightly reduced to >4 g L −1 until the end of the cultivation (Additional file 1: Figure S1). Little consumption of xylose by B. megaterium MS941 after depletion of the main carbon source was described earlier [30] and only slightly influences the amount of recombinant protein [39]. Due to the used optimized xylose-inducible promoter P xylA the D1.3 scFv concentration was found to be around tenfold enhanced compared to data published by David et al. [40].

Bioprocess scale-down and scale-up
Next, the culture volume was either reduced to 1250 μL or increased to 2 L (Table 1)  for processing a large number of parallel experiments, lab-scale stirred tank bioreactors provide more opportunities for online measuring and controlling various cultivation parameters. One of the important parameters in this study was the oxygen supply. For shake flasks and microtiter plates, oxygen input was realized via orbital shaking movements. As the ratio of gas-liquid exchange area to liquid volume strongly increased in microtiter plates compared to that provided in the shake flasks [41], a higher oxygen transfer rate (OTR) with k L a values of 150-200 h −1 was obtained. For stirred tank bioreactors, oxygen input was ensured via direct air inflow and agitation and k L a values of 75-85 h −1 were measured. However, addition of 15 mg L −1 anti-foaming agent after around 9 h of cultivation was necessary in bioreactor cultivations. Anti-foaming agents usually decreases oxygen transfer due to coalescence of bubbles and by retarding surface flow at the gas liquid interface [42,43]. The DO dropped to nearly 0% and a sudden rise of dissolved oxygen (DO) to 80% marked the transition to the stationary phase in the bioreactor cultivations (Fig. 3). In the microtiter plates, the DO declined to only 40% due to better oxygen transfer. Lower oxygen tension slowed the energy generation of the respiratory chain and the present fructose concentration caused overflow metabolism [44], indicated by the secretion of acetate during growth ( Fig. 2 and Additional file 1: Figure S1) which has been described previously [40,45]. In this study, maximal acetate concentrations of 0.45-0.55 g L −1 were attained in shake flask and bioreactor cultivations, but only 0.25 g L −1 in microtiter plate experiments. Acetate as non-preferential substrate was re-utilized after fructose depletion in shake flasks and microtiter plates. Finally, the oxygen availability also influences the amount of recombinant D1.3 scFv produced (Fig. 3). While in the bioreactor the production profile of D1.3 scFv was found to be very similar to that in shake flasks with maximal concentrations of 8 mg L −1 after 22-23 h of cultivation, in microtiter plates higher concentrations of up to 15 mg L −1 were reached already after 17-18 h. As the product was secreted in both growth and stationary phase, secretion was assumed to be mixed-growth-associated or slightly growth-independent with α = 0.28 mg scFv g −1 cells and β = 0.79 mg scFv g −1 cells h −1 for microtiter plate cultivations and α = 0 mg scFv g −1 cells and β = 0.31 mg scFv g −1 cells h −1 for bioreactor experiments (Additional file 1: Table S1). The amount of active antibody fragment remained constant also in late stationary phase or even further increased during microtiter plate cultivation (Figs. 2, 3) indicating a stable product form. Xylose-inducible gene expression can also be found in B. licheniformis [46] indicating its ability to take up xylose present in the medium. Hence, the plasmid pRBBm117 constructed for recombinant xylose-dependent D1.3 scFv secretion in B. megaterium was introduced into B. licheniformis strain MW3 [47]. Strain MW3 is a restriction-negative mutant of the type strain DSM13 in which two type I restriction modification systems were destructed enabling a better transformation and higher stability of foreign DNA compared to the wild type [47].
The growth behavior of recombinant B. licheniformis MW3 (Fig. 4) was similar for all three cultivation systems with maximal specific growth rates of 0.34-0.42 h −1 . In contrast to recombinant B. megaterium, growth was not notably reduced after addition of the inducer xylose 4-5 h after inoculation. In all three systems, fructose was consumed after 12-14 h and stationary phase was reached with about 1.5-2.0 g L −1 of cell dry weight. Less maximal biomass was formed compared to recombinant B. megaterium MS941.
In the bioreactor cultivation, DO decreased to nearly 0% at the end of the exponential growth phase (Additional file 1: Figure S2) implying oxygen limitation although fructose was still available. In microtiter plates, better oxygen transfer resulted in a less strongly decrease of DO to 50-60%. As for recombinant B. megaterium, this was accompanied by the secretion of the overflow metabolite acetate. Maximal acetate concentrations of 0.7 g L −1 were detected in bioreactors, but only 0.4 g L −1 in shake flaks and microtiter plates when the culture reached the stationary phase. Re-utilization of the secreted acetate as well as partly metabolization of xylose explains the slight biomass increase after fructose depletion in bioreactors and microtiter plates. Secretion and re-utilization of acetate [48] and the utilization of xylose as carbon source [49] were described previously.
Although growth was not affected by the addition of the inducer, slightly higher amounts of recombinant functional D1.3 scFv were secreted by recombinant B. licheniformis compared to B. megaterium. Highest product amounts were again reached in microtiter plates (17 mg L −1 after 18 h), followed by shake flasks (15 mg L −1 after 15 h) and bioreactor cultivations (13 mg L −1 after 18 h). Secretion seems to be non-growth-associated in microtiter plates (α = 0 mg scFv g −1 cells and β = 1.19 mg scFv g −1 cells h −1 ) and bioreactors (α = 0 mg scFv g −1 cells and β = 0.51 mg scFv g −1 cells h −1 ) as occurring mainly in the stationary phase (Additional file 1: Table S1). For shake flask cultivation, mixed-growth-associated secretion was observed with α = 3.286 mg scFv g −1 cells and β = 1.286 mg scFv g −1 cells h −1 (Additional file 1: Table S1). For cultivations in microtiter plates and shake flasks, extracellular scFv concentration decreases during stationary phase. Since B. licheniformis DSM13, the wild type strain of the here-used derivate MW3, is known to produce various extracellular proteases including well-studied Subtilisin Carlsberg [50], degradation of scFv by these enzymes is assumed.
Since B. subtilis was described to ineffectively utilize xylose as carbon source due to a missing specific transport system [56], the suitability of the here used and optimized xylose-inducible plasmid system had to be tested for the used strains. For this purpose, extracellular precipitated proteins were compared after xylose-induced and non-induced cultivations after 13-15 h in shake flasks via SDS-PAGE. In the medium of all three B. subtilis strains the antibody fragment D1.3 scFv, a monomeric protein with a molecular weight of approximately 27 kDa, was clearly observed for induced, but only marginally for non-induced cultivations (Fig. 5a). This indicated the enhanced production and secretion of D1.3 scFv in the presence of xylose and confirmed the correct function of the here used xylose-inducible promoter system. At the same time, the used plasmid system revealed a high segregational stability in the here tested B. subtilis strains. The application of the xylose-inducible promoter in B. subtilis was described previously but was only tested as genomically integrated system due to plasmid instabilities [57]. In addition, the amount of secreted D1.3 scFv positively correlated with the number of eliminated proteases. B. subtilis WB800N showed the highest and B. subtilis 168 the lowest concentration of D1.3 scFv (Fig. 5b). Although B. megaterium strain MS941 showed a very low protease activity [36], there was only very low extracellular overall protein and corresponding very low amount of extracellular D1.3 scFv visible compared to WB800N (Fig. 5b).

Production and secretion of recombinant penicillin G acylase in different Bacillus strains
To verify obtained results for the recombinant production of D1.3 scFv similar experiments were performed for the recombinant production of the exo-enzyme penicillin G acylase (Pac) using B. megaterium MS941, B. subtilis 168 and B. subtilis WB800N. Pac consists of a small α subunit (27 kDa) and a large β subunit (59 kDa) which are autocatalytically processed from a cytoplasmic polypeptide precursor and folded outside the cell after secretion [30,31]. For recombinant production of Pac in different Bacillus strains, the expression plasmid pALBm1 was constructed, which encodes the Pac protein from B. megaterium ATCC14945 with its original signal peptide. Transcription of the corresponding gene is controlled by the optimized xylose-inducible promoter system. All other structures of pALBm1 are identical to these of the plasmid pRBBm117 used for antibody fragment production. Cultivations of transformed B. megaterium MS941 as well as B. subtilis 168 and WB800N were performed in shake flasks with 150 mL of LB medium at 37 °C.
Growth was similar for all investigated strains. All strains reached the stationary phase after 21 h with cell dry weights of about 2 g L −1 (Fig. 6) (Fig. 6). The lowest Pac activities of 0.3 U mL −1 were observed with B. subtilis 168, which were even lower than with B. megaterium MS941 (0.4 U mL −1 ), probably showing the influence of highly active proteases, while the lack of eight proteases characterizing B. subtilis WB800N doubled the maximal activity to 0.65 U mL −1 . The observed production and secretion behavior for Pac was similar for all tested strains compared to the antibody fragment production and secretion.

Time-resolved comparison of recombinant D1.3 scFv secretion in different B. subtilis strains
Next, B. subtilis strains 168, DB431 and WB800N all transformed with pRBBm117 were individually cultivated in microtiter plates, shake flasks, and bioreactors (Table 1) investigating time-resolved growth behavior and secretion  [58] (Fig. 5b). The DO dropped to 0% 3 h (168) and 2 h (DB431) prior transition to stationary phase (Additional file 1: Figures S4, S5). Due to slower growth for WB800N, the DO of this strain only dropped down to 20% about 2 h before transition to stationary phase (Additional file 1: Figure S6) since a lower OTR into the cells was needed. The observed oxygen limitation in bioreactor cultivations resulted in the secretion of the fermentation product acetate in high amounts of 0.6 g L −1 (168), 0.7 g L −1 (DB431) and 0.9 g L −1 (WB800N) (Additional file1: Figures S4-S6). In contrast to that, culture volume reduction realized higher OTR and minimal DO values  Figure S6). Secretion and re-utilization of acetate was described previously for another B. subtilis strain [59]. Biologically active D1.3 scFv was secreted with all three recombinant B. subtilis strains in all three cultivation systems (Figs. 7, 8). Secretion was observed to be strongly (168) or mainly (DB431 and WB800N) growth-associated, since maximal concentrations were observed during or shortly after transition into stationary phase, respectively (Additional file 1: Table S1). Growth-associated secretion of a scFv with B. subtilis was described earlier with much lower product concentration although using a protease deficient strain [60]. In this study, maximal formed and secreted D1.3 scFv concentrations significantly differed depending on the used B. subtilis strain and the employed cultivation system. Maximal D1.3 scFv amounts in shake flasks were 30 mg L −1 (168, after 12-13 h), 50 mg L −1 (DB431, after 14 h) and 120 mg L −1 (WB800N, after 16-18 h), respectively, while for B. subtilis DB431 and WB800N, slightly higher concentrations were observed in microtiter plates (60 and 130 mg L −1 ) due to better OTRs. For bioreactors, lower maximal D1.3 scFv concentration of 20 mg L −1 (168), 30 mg L −1 (DB431) and 70 mg L −1 (WB800N) were obtained, which might be due to the limited supply of oxygen or a negative effect of the added anti-foaming agent on product formation or secretion [61]. Compared to B. megaterium MS941 and B. licheniformis MW3, higher D1.3 scFv concentrations were achieved in less cultivation time with all three B. subtilis strains. Bacillus subtilis 168 did not only secrete the smallest amount of D1.3 scFv among the examined B. subtilis strains but its extracellular scFv concentration also strongly decreased during stationary phase for all cultivation systems. Especially in stirred tank bioreactors, scFv concentration dropped dramatically fast in stationary phase of cultivation, resulting in a quite narrow and small concentration profile. Here, degradation by proteases was strongly assumed, since B. subtilis 168 is known to produce and secrete high amounts of different proteases, especially at the end of growth phase [62]. Apparently, the absence of six proteases in B. subtilis DB431 resulted in higher and more constant scFv titer profiles, although degradation of scFv during stationary phase still occurred (Fig. 8). Finally, B. subtilis WB800N, lacking eight proteases, secreted the highest scFv amounts for all three cultivation systems and no significant degradation of scFv occurred in any cultivation system during stationary phase clearly indicating the positive influence of proteases deficiency on the recombinant protein production and secretion in this B. subtilis strain.

Space-time-yield (STY) of D1.3 scFv secretion in Bacillus
For overall balancing and evaluation of the production and secretion of D1.3 scFv, the space-time-yield (STY) was calculated as the proportion of scFv concentration and corresponding cultivation time (Fig. 9).

Discussion
In this study, different strains of the genus Bacillus were shown to efficiently produce and secrete the recombinant antibody fragment D1.3 scFv as well as penicillin G acylase Pac in three different cultivation systems. For this, all strains were successfully transformed with the identical plasmid encoding D1.3 scFv and Pac, respectively, together with an optimized B. megaterium xylose-inducible promoter system and B. megaterium originated signal peptides [63]. This clearly shows the broad interspecies application of an expression plasmid in Bacillus, which was not investigated so far. Production and secretion of scFvs and scFabs with B. megaterium were examined before using its native xylose-inducible promoter system. With this system, 390 µg L −1 of anti-CRP scFv [64], 410 µg L −1 of D1.3 scFv [33] and 3.5 µg L −1 of the more complex D1.3 scFab were produced and secreted in their active forms by B. megaterium strain MS941 [64] deficient in the main extracellular protease NprM lacking around 98% of the overall extracellular protease activity [36]. When compared to the production and secretion by a B. megaterium strain additionally deficient in xylose metabolization [30], D1.3 scFv secretion was enhanced to 700 µg L −1 in shake flasks and even 3 mg L −1 in 3 L bioreactors [40,65]. With E. coli, about 1-2 mg L −1 of D.13 scFv were recovered as soluble protein in the culture supernatant and as periplasmic fraction [66], showing the inefficient production in Gram-negative bacteria.
For B. licheniformis, no secretion of any scFv was described so far. For the here-used strain B. licheniformis MW3, scFv concentrations in the same range as for B. megaterium were achieved. Among the tested B. subtilis strains, highest amounts of active D1.3 scFv were secreted with the strain WB800N deficient in eight proteases. WB800N showed the highest product yield which was 8.7-fold higher than the best B. megaterium   described so far. In the literature, B. subtilis strains, mostly mutants of strain 168 with deletions of one or more protease genes, show very low overall extracellular protease activity leading to improved yields of various products compared to the wild type strain [21]. Secretion of up to 15 mg L −1 with B. subtilis was described before [53,60,67]. The three-protease-deficient B. subtilis WB30 was reported to secrete 3.5 mg L −1 and the sixprotease-deficient WB600 5 mg L −1 of an scFv against digoxin [60]. With coproduction of intracellular and extracytoplasmic molecular chaperones, scFv titers of 12 mg L −1 were observed [67]. With B. subtilis WB700N, a seven-extracellular-protease-deficient derivate, secretion of scFv against fibrin even failed [53]. However, the combination of enhanced coproduction of molecular chaperones and additional inactivation of the cell-wall-associated protease WprA in the strain WB800HM[pEPP] resulted in scFv titers of up to 15 mg L −1 [53]. In addition, correlation of increased stability of the produced scFv with less secreted proteases was observed comparing B. subtilis 168, WB30 and WB600 [60]. In accordance with that, in this study higher and more constant scFv titer profiles were observed for the protease-deficient B. subtilis strains DB431 and WB800N.
Apart from B. megaterium and B. subtilis, production of extracellular recombinant antibody fragments has been reported only for B. brevis and Lactobacillus spp. With B. brevis, secretion of about 100 mg L −1 of a Fab against human urokinase-type plasminogen activator was reported [68]. Lactobacillus spp. were used to secrete 6 µg L −1 of Guy's 13 scFv against a streptococcal adhesion molecule [69], 410 µg L −1 of codon-optimized antiviral 3D8 scFv [70] and 1 mg L −1 of an of llama heavy-chain (VHH) antibody fragment against rotavirus [71].
In contrast to Bacillus-based production systems mammalian cell cultures produce antibodies at g L −1 scale. However, much longer cultivation times of up to 3 weeks are needed. For example, using a human cell line 400-500 mg L −1 of IgG were produced within 10 days, corresponding to a STY of 1.66-2.08 mg L −1 h −1 [72]. IgG concentrations of 2.64 g L −1 (in 13 days) were realized with NSO cells, which corresponds to a STY of 8.46 mg L −1 h −1 [73]. For CHO cells up to 12.8 mg L −1 h −1 of IgG were described [74][75][76]. Since we observed maximal secretion with the tested Bacillus strains in only 9-21 h, the resulting STY of up to 8 mg L −1 h −1 are in the same order of magnitude as described for mammalian cells, showing the ability to compete with these cultivation systems for scFv-secretion. However, cultivation costs are characteristically smaller for the bacterial production systems.
Recombinant production of penicillin G acylase Pac from B. megaterium ATCC14945 in other B. megaterium [30] and B. subtilis [77][78][79] strains was shown previously, using various cultivation systems and mainly complex media. Here, we used the production of this enzyme for the confirmation of the results obtained for scFv-production and secretion. In agreement, the deficiency for eight proteases in B. subtilis WB800N led to a doubling of the maximal activity while the presence of all natural protease clearly reduced the amount of Pac in the medium, validating the assumptions made for D1.3 scFv production in protease-deficient B. subtilis strains. Overall, similar results were obtained in comparison to the recombinant secretion of D1.3 scFv, clearly confirming the general applicability of the introduced systems.

Conclusions
Different tested strains of the genus Bacillus provided high potential microbial production systems to synthesize and export challenging heterologous proteins like antibody fragments and high molecular weight proteins at high titers in different cultivation systems: a microtiter plate approach (1250 μL culture volume), uncontrolled shake flasks (150 mL culture volume) and a controlled laboratory-scale stirred tank bioreactor (2 L culture volume). The employed plasmid originally optimized for the recombinant protein production in B. megaterium can also efficiently be used for B. licheniformis and B. subtilis strains, which demonstrates its interspecies application. Kinetics of D1.3 scFv production were successfully described with the Luedeking-Piret model for all tested Bacillus strains and cultivation systems. Highest production and secretion of the antibody fragment was attained in microtiter plates and shake flask because of their better oxygen transfer compared to the stirred tank bioreactor approach. B. subtilis WB800N, deficient in eight proteases, secreted 130 mg L −1 (STY of 8 mg L −1 h −1 ) of the recombinant antibody fragment D1.3 scFv, which is the highest concentration and STY of a recombinant antibody fragment for a Grampositive bacterium described so far. Finally, interspecies application of the plasmid system were also shown for the secretion of a recombinant heterodimeric protein, the penicillin G acylase Pac.

Strains and plasmids
All strains and plasmids used in this study are listed in Table 2. For cloning procedures, E. coli strain DH10B acted as host (Life Technologies, Carlsbad, USA). B. megaterium strain MS941 is a derivate of the wild type strain DSM319 with a defined deletion in the gene encoding the major extracellular protease NprM [36,80]. B licheniformis strain MW3, a derivate of the type strain DSM13, has defined deletions in genes encoding two type I restriction modification systems (HsdR1, HsdR2) to improve transformability and recombinant DNA stability [47]. B. subtilis 168, the most widely biotechnologically used B. subtilis strain, is originated from B. subtilis Marburg via X-ray-mutagenesis. It is tryptophan-requiring auxotroph [51]. B. subtilis strain DB431 is deficient in four extracellular and two intracellular proteases [52], while strain WB800N lacks six extracellular, one cell-wall-associated and one intracellular protease [53,54]. For generation of the here used expression plasmid pRBBm117, the D1.3scFv gene fused to the coding sequences of the signal peptide of the B. megaterium lipase A [63] at its 5′-end, of a His 6 -tag at its 3′-end and of a terminator structure were liberated from plasmid pEJBmD1.3scFv [33] using the restriction enzymes BsrGI and AgeI. The fragment was cloned into similarly cut p3STOP1623hp [32] resulting in pRBBm117. For generation of expression plasmid pALBm1, the pac gene encoding the penicillin G acylase was amplified from genomic DNA of B. megaterium ATCC14945 using the primers pac_fw (5′-TACATATGTACAATGAAGACGAA-GTGGCTAATATCA-3′) and pac_rv (5′-TATCAGAGCT-CATCAATAGTATAGGCTC-3′). The pac gene including the native signal peptide was cloned into the shuttle vector p3STOP1623hp [32] using SacI und BsrGI.

Isolation of genomic DNA from B. megaterium ATCC14945
For preparation of chromosomal DNA from B. megaterium ATCC14945, the JETquick Plasmid Miniprep Kit (Genomed, Löhne, Germany) was used according to the manufacturer's instructions with some modifications. Cells from a cryogenic glycerol stock were grown overnight at 37 °C and 130 min −1 in 50 mL of LB medium in baffled flasks. Cells of 6 mL culture suspension were harvested and suspended in 250 µL TE(H) buffer (150 mM Tris-HCl, 20 mM EDTA, pH 8) before 20 µL of lysozyme solution (20 g L −1 lysozyme in TE(H) buffer) were added. Cells were incubated for 10 min at 37 °C and 750 min −1 before adding the next buffer. Before elution of DNA, the column was washed with 500 mL of GX buffer and 500 µL of G4 buffer.

Transformation and strain maintenance
Unless mentioned here, molecular biology methods used in this study were described previously [81]. Plasmid DNA for the transformation of B. megaterium MS941, B. licheniformis MW3 and B. subtilis 168 was prepared from recombinant E. coli while plasmid DNA for transformation of B. subtilis strains DB431 and WB800N was prepared from B. subtilis 168 already carrying the plasmid pRBBm117. All DNA preparations were done by using the QIAgen Spin Miniprep Kit (QIAgen, Hilden, Germany) according to the manufacturer's instructions with some modifications. For the preparation of plasmid DNA from Gram-positive B. subtilis, lysozyme (1 mg mL −1 ) was added to the resuspension buffer for suspending cells of 50 mL culture (OD 600 of 1). Cells were then incubated for 10 min at 37 °C before adding the next buffer.
The protoplast transformation of B. megaterium was performed as outlined previously [82]. The protoplast transformation of B. licheniformis was performed according to Waschkau et al. [47]. All B. subtilis strains were transformed via induced natural competence. All strains were stored at −80 °C in a 30% (v/v) glycerol solution in cryogenic vials. For preparing cryogenic cultures, 50 mL of LB medium [81] with 10 mg L −1 of tetracycline for plasmid-carrying recombinant strains were inoculated with cell material from a LB agar plate and incubated at 37 °C and 230 min −1 . During exponential phase (OD 600 = 1-2), 32.5 mL of the cell suspension were mixed with 17.5 mL of glycerol [86% (v/v)], immediately frozen in 1 mL aliquots in liquid nitrogen and stored at −80 °C.

Cultivation of recombinant strains
The minimal medium used in this study for production of recombinant D1.   B. subtilis WB800N). These pre-cultures were grown in 500 mL baffled flasks at 37 °C and 130 min −1 for 18 h. Main cultures were inoculated with a pre-culture volume adjusting to an OD 600 of 0.1. Inoculation culture was centrifuged (1500×g, 5 min, room temperature) before suspended in fresh minimal medium. Expression of the D1.3 scFv was induced by addition of 5 g L −1 d-xylose at an OD 600 of 0.3-0.4.
For shake flasks cultivation for production of recombinant D1.3 scFv, 150 mL of minimal medium were incubated in 500 mL baffled flasks (3 baffles) at 37 °C and 130 min −1 (50 mm orbital shaking diameter). Shake flask cultivations were carried out in triplicate. For bioreactor cultivation, 3 L bioreactors (Applikon, Schiedam, The Netherlands) with two six-bladed Rushton turbine impellers were used. Cultivation was performed with 2 L of minimal medium. Temperature (37 ± 0.1 °C), aeration rate (3.0 L min −1 , 1.5 vvm), agitation speed (500 min −1 ) and pH value (pH 6.5 ± 0.1, adjusted with 2 M HCl and 2 M NaOH) were automatically kept constant. Dissolved oxygen (DO) was measured online. Bioreactor cultivations were carried out in duplicate (strains MS941, MW3, 168) and triplicate (strains DB431, WB800N), respectively. To prevent foam formation, anti-foaming agent (Ucolub ® , FRAGOL GmbH & Co. KG, Mühlheim/Ruhr, Germany) was added when required. For microtiter plate cultivation, 1250 µL of minimal medium was transferred to each well of a 48 Well FlowerPlate ® (m2p-labs, Baesweiler, Germany) with optodes for pH and DO. Plates were covered with an adhesive gas-permeable membrane (Thermo Scientific, Dreieich, Germany) and incubated in a BioLector ® unit (m2p-labs, Baesweiler, Germany) at 37 °C and 1000 min −1 with an orbital shaking diameter of 3 mm. The calibration values for the measurement of pH and DO were obtained from calibration data sheets provided by m2p-labs. According to manufacturer (m2plabs, Baesweiler, Germany), the oxygen transfer rate in 48 Well FlowerPlates ® comes to 35-40 mmol L −1 h −1 for the used conditions (see above).
Cultivation of all plasmid strains for the secretion of recombinant penicillin G acylase was performed in shake flasks as described for the production of D1.3 scFv above but took place in LB medium [81]. Main cultivations were performed in duplicates. Samples of cultivation supernatant were taken for penicillin G acylase activity assay as described below. For this, 1 mL cell suspension was centrifuged (9000×g, 5 min, 4 °C), supernatant was filtered (polyvinylidene fluoride, 0.2 μm pore size, Roth, Karlsruhe, Germany) and stored at 4 °C before use.

OD 600 and cell dry weight
Cell concentration was determined as optical density at a wavelength of 600 nm using a Libra S11 Visible Spectrophotometer (Biochrom GmbH, Berlin, Germany). Cell dry weight (CDW) was measured via gravimetric analysis. Here, 20 mL culture volume (duplicate) were pelleted (1500×g, 15 min, 4 °C) in Corex glass tubes, washed twice with distilled water to remove salts and dried at 80 °C for 48 h. The relationship between CDW and OD 600 was determined as CDW [g L −1 ] = a [g L −1 ] · OD 600 [-] with coefficient a equals 0.223 ± 0.007 g L −1 (B. megaterium MS941), 0.40 ± 0.01 g L −1 (B. licheniformis MW3) and 0.337 ± 0.002 g L −1 (all B. subtilis strains used in this study), respectively.

High performance liquid chromatography (HPLC) analysis
Sugars (xylose, fructose) were quantified by an HPLC system (HitachiElite LaChrom, Krefeld, Germany) equipped with a Metacarb 87 °C column (Varian, Palo Alto, CA, USA) as stationary phase and Millipore H 2 O as mobile phase at 0.6 mL min −1 and 85 °C with a retention time of approximately 11.7 min (xylose) and 13.4 min (fructose). Detection was performed using a refractive index detector (RI) detector. Acetate were also measured using an HPLC system (Hitachi Elite LaChrom, Krefeld, Germany) equipped with an Aminex HPX 87 H column (Biorad, Hercules, CA, USA) as the stationary phase and 12.5 mM H 2 SO 4 as mobile phase at 0.5 mL min −1 and 45 °C with a retention time of approximately 18.5 min. Detection was performed using a RI detector.
On each plate, a calibration standard of purified D1.3 scFv was applied. Standards were purified from culture supernatant with immobilized metal affinity chromatography (IMAC) and quantified by densiotometric analysis as described below. Samples for ELISA were measured in adequate dilutions (with PBS) and quantified according to the standard via five-parameter logistic analysis.

Immobilized metal affinity chromatography (IMAC), sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and densiotometric analysis
Recombinant D1.3 scFv was purified via IMAC. For this purpose a gravity flow column (BIORAD, Hercules, California, USA) was loaded with 1 mL Chelating Sepharose Fast Flow (Amersham Biosciences, Freiburg, Germany), washed with 10 mL H 2 O, incubated with 2 mL 100 mM NiSO 4 for 10 min and washed with 5 mL H 2 O followed by 5 mL washing buffer (50 mM NaCl, 20 mM Tris-HCl pH 8.6). The nickel charged Sepharose was transferred to 50-100 mL of culture supernatant containing 7 mM of imidazole and mixed for 90 min at 4 °C. The Sepharose was pelleted by centrifugation (1500×g, 5 min, 4 °C) and washed with 3 mL washing buffer. Elution was performed with 3 × 1 mL elution buffer (50 mM EDTA in washing buffer). 18 µL of each fractions were analyzed by SDS-PAGE analysis (12% of acrylamide) as described before [83]. Gels were stained with SimplyBlue ™ SafeStain according to the manufacturer's instructions for the microwave procedure (Thermo Fisher Scientific Inc., Waltham, Massachusetts, USA). Band visible on the stained SDS PAGE corresponding to recombinant D1.3 scFv in the elution fractions from IMAC were quantified by densitometric analysis (ImageJ 1.48v, National Institute of Health, Bethesda, USA) related to a BSA standard (10-100 mg L −1 ).

Precipitation of extracellular proteins
Extracellular proteins of B. megaterium and B. subtilis strains cultivated in shake flasks were precipitated by incubating 1.5 mL culture supernatant with 660 mg (NH 4 ) 2 SO 4 for 2 h at 4 °C in an overhead shaker. After centrifugation (9000×g, 30 min, 4 °C), precipitated proteins were suspended in 80 µL of 8 M urea supplemented with 27 µL of 4-times SDS loading buffer. 15 µL of this were analyzed by SDS-PAGE as described above.

Determination of the volumetric mass transfer coefficient (k L a) in bioreactors
k L a value in bioreactors was determined with sulfite oxidation method, which based on oxidation of Na 2 SO 3 to Na 2 SO 4 by oxygen. Measurement was performed under cultivation conditions (described above) without cells. 40 mg L −1 of Co(NO 3 ) 2 were added as catalyst. DO was monitored online with an oxygen sensor (Visiferm DO 225, Hamilton, Höchst im Odenwald, Deutschland). Calibration to 100% was performed under oxygen saturation condition (DO max = 6.784 mg L −1 at 37 °C). Complete oxygen depletion was achieved by addition of 10 mL of a 200 g L −1 Na 2 SO 3 stock solution [84]. As soon as the oxidation was completed the DO increased again. Plotting ln(DO max − DO) against time, the linear slope was used to determine the k L a value.

Kinetics analysis of D1.3 scFv production
The Luedeking-Piret model was used for modeling the D1.3 scFv production and secretion [85]. The observed experimental product profile from cultivation start to the time point of maximal D1.3 scFv concentration was approximated by the numerical integration of the Luedeking-Piret equation here, t n and t n − 1 are the present and previous time steps of integration. P n and P n − 1 are concentrations of the product D1.3 scFv and X n and X n − 1 the cell dry weights at the present and previous time steps. The growthassociated Luedeking-Piret constant α [mg scFv g −1 cells ] equals the biomass-related product yield coefficient Y P/X for D1.3 scFv formation. The non-growth-associated Luedeking-Piret constant β [mg scFv g −1 cells h −1 ] represents the specific production rate for specific growth rate µ = 0 h −1 . These biological reaction parameters were set so that the coefficient of determination R 2 was maximal, representing the goodness of fit of the model concentrations of D1.3 scFv compared to the experimental concentrations. Calculation was performed with the Generalized Reduced Gradient algorithm of the addin program Solver for Microsoft Excel 2010 (Microsoft, Redmund, WA, USA).